January 2021 No: 1324 - University of Warwick

Choosing the narrative :
 The shadow banking crisis in the light of Covid

 Marcus Miller

 (This paper also appears as CAGE Discussion paper 534)

 January 2021 No: 1324

 Warwick Economics Research Papers

ISSN 2059-4283 (online)
ISSN 0083-7350 (print)

Choosing the narrative: the shadow banking crisis in light of Covid
 Marcus Miller1

 University of Warwick

 January, 2021

 Abstract
Could experiencing a health pandemic aid in understanding the nature of financial crisis?
It might, for example, help to discriminate between different narratives that claim to do
so. In this spirit, two influential accounts of the near-collapse of shadow banking in the
US financial crisis of 2008 are analysed: one developed by Mark Gertler and Nobuhiro
Kiyotaki and the other presented by the Financial Crisis Inquiry Commission of the US
Congress.

Using a common two-sector framework, key features of these contrasting accounts of
the market for banking services are presented, along with their corresponding diagnoses
of what precipitated financial crisis. To see what the experience of Covid might imply
about their relative credibility, four aspects of the current pandemic are considered:
how it began from a small biological shock; how it gets spread by contagion; the
significance of externalities; and how it may end with a vaccine. But the reader is left to
form his or her own judgement.

Keywords: shadow banking, rating agencies, equity constraints, technology and news
shocks, bank runs, epidemic and social contagion.

 JEL Classification: G01, G24, G41, Y8, Z13

Corresponding Author: Marcus Miller, Professor Emeritus, Department of Economics,
University of Warwick, Coventry, CV47AL, UK. Email: marcus.miller@warwick.ac.uk

1. Introduction

In “The Jungle Tale and the Market Tale”, one of his Economic Fables, Ariel Rubinstein
(2012) offers two accounts of how an economy might allocate resources – by the ‘iron
hand’ of relative strength; or by the ‘invisible hand’ of competitive equilibrium. In both
cases, he promises, ‘I will follow economic tradition and demonstrate the ideas via
models – tales or fables2 … But I will not use formal language, which would make the

1
 Acknowledgements: While taking full responsibility for views expressed, the author would like
to acknowledge his indebtedness to joint work with Lei Zhang, CEPR DP No. 13834 in particular;
and to thank both seminar participants at the University of North Bengal at Siliguri and colleagues
Peter Hammond, Alex Karalis, Robert Mackay, Federico Rossi and Marija Vukotic for valuable
comments and suggestions. The support of the Economic and Social Research Council (ESRC) is
gratefully acknowledged, via the Rebuilding Macroeconomics Network (Grant Ref:
ES/R00787X/1).
2
 ‘The word model sounds more scientific than the word fable or tale, but I think we are talking
about the same thing.’ Rubinstein (2012, Chapter 1, p.16).

 1

models easier to understand for the few who are familiar with this language, but would
pose an impenetrable barrier for all the rest.’

It is in this spirit that two influential accounts of near-collapse of shadow banks in US
financial crisis of 2008 are examined here. This is not intended, however, as an
intellectual exercise of comparing market and non-market mechanisms for allocating
resources. Rather it is to see what we might learn from the health crisis - by comparing
two influential but sharply contrasting accounts of the market for banking services and
of what went wrong in the financial crisis; then seeing what the current experience of
the Covid pandemic might imply about their relative credibility.

Four aspects of the current pandemic are considered: how it began from a small
biological shock; how it gets spread by contagion; the significance of externalities; and
how it may end with a vaccine. But the reader is left to form his or her own judgement.

Contrasting Narratives

One of the narratives to be discussed has been developed by distinguished academics,
Mark Gertler of Columbia University and Nobuhiro Kiyotaki of Princeton University, in a
series of technical papers appearing in leading economic journals and handbooks since
2015. This, the Scholars’ Tale, holds that US shadow banks3 provided highly efficient
financial intermediation – with lower spreads between expected asset returns and
borrowing costs than commercial banks, for example. Following hard on the heels of an
unanticipated negative productivity shock, however, the financial sector suffered a
systemic ‘bank run’, attributable to a ‘sunspot’ – a metaphor for some random, pay-off
irrelevant, extrinsic trigger. In short, shadow banks are portrayed as efficient but fragile
institutions exposed to extraneous random shocks for which they bear no responsibility.

A starkly different picture is painted in the majority report of the Financial Crisis Inquiry
Commission (FCIC) set up by US Congress to investigate the origins and course of the
crisis. It stresses the significance of information asymmetries, and the disingenuous role
played by the Credit Rating Agencies (CRAs) in particular. Early in the Final Report, FCIC
(2011, p.xxv) one reads:

 The three credit rating agencies were the key enablers of the financial meltdown.
 The mortgage-related securities at the heart of the crisis could not have been
 marketed and sold without their seal of approval. … Their ratings helped the
 market soar and their downgrades through 2007 and 2008 wreaked havoc across
 markets and firms. From 2000 to 2007 Moody’s rated nearly 45,000 mortgage
 related securities as triple-A. This compares with six private sector companies in
 the United States that carried this coveted rating in early 2010.

3
 These are investment or wholesale banks that operated “without the security of insured
deposits that never run, without effective regulatory constraints on their leverage and without
the ability to turn to the Fed if their financing evaporated.” Bernanke et al.(2019, p.23)
 2

Specifically, the FCIC alleges that competition for business between different rating
agencies enabled investment banks to secure unrealistically optimistic ratings for assets
that they held and sold; and that this overvaluation played a key role in the ensuing
crisis.

This, the Congressional Tale, later received academic support from George Akerlof and
Robert Shiller as a case study showing how those with better information can profit at
the expense of the less well-informed4. In the push to provide credit to subprime
borrowers, they argue, investment banks, protected by the shield of limited liability,
were tempted to assemble highly-rated but very risky securitised assets for sale to other
investors and to hold in their own portfolios. Crisis arrived when such practices came to
light. Unlike the ‘sunspot’ model of banking crisis endorsed by Gertler, Kiyotaki and co-
authors (hereafter G&K for brevity), the account of Akerlof and Shiller is concerned with
the formation of beliefs in a setting of imperfect information.

What is to follow

Could these narratives of crisis not first be compared one against the other – in a type of
intellectual beauty contest where contestants seek to be most persuasive? To this end,
the stage is set in the next section by outlining a basic two-sector framework, where
shadow banks using borrowed funds compete with non-bank ‘direct’ lenders, leading to
a market equilibrium with substantial intermediation. In section three comes the beauty
contest, where each narrative offers its own distinctive twist to the framework, and
shows how this can lead to banking collapse. In section four, the light that experience of
the current medical epidemic may throw on these competing narratives of financial
crisis is considered. Section five concludes.

2. A basic framework to set the scene

The basic framework to be used involves competition between ‘direct’ lending and
‘intermediated’ lending by shadow banks to satisfy the needs of the ultimate borrowers.
This is shown schematically in Fig. 1 from Shin (2010, p.30), where the securities issued
by the ultimate borrowers (on the left) are taken up by risk-neutral shadow banks,
relying heavily on borrowed funds, and other non-bank holders (on the right) who use
their own resources and are risk averse.

4
 Chapter 2 of Akerlof and Shiller (2015)
 3

Banks
 Debt
 Intermediated Households
 (Active claims
 credit
 End-user investors) (Passive
 borrowers investors)

 Directly granted credit

 Fig. 1. Intermediated and directly granted credit
 Taking the amount of risky assets issued by end-user borrowers to be fixed, the market
outcome proposed by Shin himself is illustrated5 in Fig. 2 below. Payoffs for the risky
asset are assumed to have a uniform distribution between the upper and lower bounds
shown, with mean return q and downside risk . The combined demand by both sectors
determines the price of risky assets, p, measured on the vertical axis.

The demand curve for banks, assumed to maximise the expected value of their holdings
subject to an equity constraint to prevent risk-shifting, is measured from the left-hand
axis. Given the specific Value at Risk (VaR) rule that ‘equity, , must cover all downside
risk to which the portfolio is exposed’, bank demand for ≤ is in fact shown by the
rectangular hyperbola passing through K (asymptotically approaching the vertical axis
and the lower bound of risky payoffs as = /( − ( − )). There is, of course, no
demand at prices standing above the mean expected payoff.

Graphically, the kink at (q, /z), labelled K, indicates the volume of assets with downside
risk covered by equity when price is at the mean; and the downward slope thereafter
indicates, not risk aversion, but the effect of adhering to the VaR rule6. The demand
from the other, risk-averse sector, measured from the right-hand axis, increases in
proportion as the price falls below q (i.e. = ( − ) with slope, −1/ , flatter the
higher the tolerance for downside risk7). Market-clearing equilibrium is at A, where the
schedules intersect at price 0 .

5
 As in Shin (2010, Chapter 3, Fig. 3.2)
6
 which, for given equity e, allows for greater holdings the less risk exposure per unit.
7
 As formally the parameter = 3 / 2 , where τ is a measure of risk tolerance taken as fixed, Shin
(2010,p. 31.)
 4

p
 Upper bound
 q+z

 K
 q q
 Bank Demand curve η
 0
 A
 q-z
 Lower bound
 η−1 D

 0 0
 Supply of risk assets, 1

Figure 2. The price of risky assets determined by market-clearing

Since the equilibrium price lies ‘within the band’ between and − , banks will
expect to make a positive Return on Equity (ROE), especially if no interest is paid to
creditors who bear no risk8 (given that equity covers the total risk exposure on the
balance sheet). That shadow banks can hold risky assets while issuing money-like
liabilities is the so-called alchemy of finance – sustained by shadow bank shareholders
shouldering substantial risk.

The framework just described effectively assumes Common Knowledge: so, as portfolio
managers in both sectors know the true values of parameters such as and (and know
that others do), there is strictly speaking no role for credit rating agencies. In the
concluding chapter of Risk and Liquidity, however, Shin indicates that, in practice, things
are not so straightforward. For credit ratings are widely used; yet, one is warned, ‘heavy
reliance on credit rating agencies … is misguided. They are unregulated and the quality
of their risk estimates is largely unobservable.’ Shin (2010, p.171) Taking our cue from
this remark, we turn next to the perspective of the FCIC with its focus on issues of
asymmetric information.

8 − 
 so = ( − ) = = 
 
 5

3. A beauty contest: the narratives compared

3.1 The Congressional Tale: of fooling, revelation and panic

As a preliminary, we note that, for Bolton and Dewatripont (2005, p. 175), a ‘central
result in the literature on disclosure of information is that when certification is costless,
there is full disclosure of information in equilibrium under very general conditions.’ So it
might be thought that the existence of well-established credit rating agencies – and the
active role that they play in US securities markets9 - could have ensured that issues of
information asymmetries were largely finessed - with conservative banks borrowing at
low rates of interest compared with risky banks paying much higher rates, for example.
But if, as alleged by the FCIC Report, ratings are the outcome of competition for business
between unregulated oligopolists - no longer run as partnerships but as profit
maximising corporations10 - then full disclosure of information is unlikely. Flattering
assessments of product quality, which promote the profits of clients who pay for the
ratings, are a tempting way of keeping them on the books; and, the Commission
concluded, “the rating agencies placed market share and profit considerations above the
quality and integrity of their ratings. “ FCIC (2011, p. 212)

A ‘fooling’ equilibrium that unravels

By its own account, the Congressional Commission ‘interviewed more than 700
witnesses, held 19 days of public hearings, and examined millions of pages of
documents’; and its majority Report runs to some 400 pages. Though no formal model is
offered, the tale that is told may be illustrated by reinterpreting the basic framework
spelled out above.

 What the tale implies, however, is that an equilibrium in the market for risky securities
seemingly based on Common Knowledge should be treated with caution. If behaviour
was based on false certification, it will instead lead to a type of ‘fooling equilibrium’11
where creditors who lend to shadow banks holding highly-rated subprime securities
underestimate the true risks being taken with their funds, and fail to realise their own
exposure until too late.

Why banks should wish to get the risky assets they hold - and assemble for sale - over-
rated is not far to seek: it is a way of exploiting the shield of limited liability. For over-
rating is a way of working around the VaR rules, shielding shareholders by exposing

9
 ‘In 2007 Moody’s Standard and Poor’s and Fitch [the leading US credit rating agencies]
employed about 14,000 people between them, covering 95% of the rating market. Although the
rating agencies call their ratings mere ‘opinions’, these opinions have legal power as far as
pension funds, boards of trustees of charities and other institutions are required by law to hold
only investment grade securities’. Sinn(2010, pp.127,8)
10
 Moody’s Corporation went public in 2000; and, ‘after the public listing, the company culture
changed – it went “from [a culture] resembling a university academic department to one which
values revenues at all costs”.’FCIC (2011, p. 207)
11
 In the terminology of Akerlof and Shiller (2015).
 6

creditors to significant downside risk. With more expected profits to be made on the
upside, but extra losses transferred to creditors on the downside, hidden risk raises the
expected ROE for the banks – assuming creditors also believe the rosy ratings.12

It is the contention of Goodhart and Lastra (2020), indeed, that “the limited liability of
equity holders is by far the biggest source of moral hazard and risk shifting in a capitalist
economy”; and the diagram from their paper, see Fig. 3, conveys this critique most
cogently.13 Even though there is a safe investment that gives a certain return of C to
equity holders, they can expect to do better if the bank invests in risky securities - as, for
example, one with the same mean and two equi-probable payoffs, A which is highly
profitable and D where losses will drive the bank into insolvency. With liability limited to
e, equity invested, losses on the downside are shared with the creditors, as shown; so
the expected payoff to equity rises from to ∗ - but only so long as the creditors are
‘fooled’.

 Payoff for equity
 Retourn to equity
 Not to scale

 A

 C*

 C
 Portfolio payoff
 Bank insolvent B
 -e
 Creditor loss

 D

Fig.3 With limited liability: ‘gains we keep, losses we share’.

If ‘at the length truth will out’, then such fooling equilibria will be liable for correction.
What to expect when the true quality of the risk assets is revealed – i.e. after what

12
 If not, creditors can raise the rate they charge for funding, as in Adrian and Shin (2014).
13
 It is for a two-point, equi-probable distribution of payoffs, however, rather than the uniform
density of Fig. 2. But note that, even without a safe security, creditors would still achieve safety if
the bank equity covers the downside risk as in Fig.2 – unless, unseen by them, the ROE is raised
by widening the spread of returns.
 7

Bernanke (2018) describes as ‘bad news’ arrives? The basic framework – suitably
reinterpreted - can provide a simple account of the Congressional Tale.

If the initial equilibrium was based on an under-estimate of the actual downside risk,
then the immediate effects of realising the truth are shown in Fig. 4. As before,
equilibrium is determined by market clearing at A, but this is now interpreted as a
‘fooling’ equilibrium, , based on an underestimate of the riskiness of the assets
involved. The increase of perceived risk when the truth is revealed will make these
assets less attractive to both sectors. For passive investors the rise in downside risk
(from to > ) makes their demand schedule steeper, as indicated by the anti-
clockwise movement in the figure. The demand curve for shadow banks, subject to a
binding VaR constraint, shifts to the left (from K at / to K’ at / at the top of the
figure) as the unit risk increases to its true value; and becomes steeper as the lower
asymptote moves down to − ), the correct lower bound. Hence equilibrium will, on
impact, move from A to B as shown, with a fall in the price but not much trading of
assets14.

 p
 ′ q
 q

 Bank Demand A
 Non-bank Demand
 η−1 B Demand η
 C
 q-z
 B

 q- 
 Correct lower bound
 D

 0 Supply of risk assets, 1

Fig. 4. Effect of unanticipated Bad News: on impact and subsequently

This reduction of demand is not the end of the story, however. With assets ‘marked-to-
market’ at lower prices, the fall in shadow bank equity permits less asset-holding as the
VaR constraint binds tighter. This endogenous adjustment of bank equity will amplify the
effect of Bad News, with investment banks losing market share as equilibrium shifts
along the demand curve for passive investors to a point like C.

14
 As broadly illustrated in calibrations reported in Miller and Zhang (2019), CEPR DP. 13834.
 8

Will these mark-to-market effects lead to insolvency of shadow banks? Not if, as shown,
the falling price still lies within the range for which equity provision was made (i.e. C lies
above p = q-z). But this is without taking into account the reaction of creditors.

Creditor panic

In his account of the crisis, Ben Bernanke (2018) describes how creditors respond to bad
news:

 Before the crisis, investors (mostly institutional) were happy to provide wholesale
 funding, even though it was not government insured, because such assets were
 liquid and perceived to be quite safe. Banks and other intermediaries liked the
 low cost of wholesale funding and the fact that it appealed to a wide class of
 investors. Panics emerge when bad news leads investors to believe that the “safe”
 short-term assets they have been holding may not, in fact, be entirely safe. If the
 news is bad enough, investors will pull back from funding banks and other
 intermediaries, refusing to roll over their short-term funds as they mature. As
 intermediaries lose funding, they may be forced to sell existing loans and to stop
 making new ones.
How to factor in such behaviour? For commercial banks, creditor panic takes the form
of a bank run with customers withdrawing deposits; but for non-deposit-taking
institutions like shadow banks, the funding squeeze shows up instead as a sharp rise in
the ‘haircut’ imposed on collateral for repo borrowing15. As Shin (2010, p. 144, 145)
notes, ‘fluctuations in the haircut largely determine the degree of funding available to a
leveraged institution … [and] times of financial stress are associated with sharply higher
haircuts’.

Just how much higher Shin illustrates by the rise in haircuts from April 2007 before the
crisis to August 2008 in its midst, so ‘a borrower holding AAA-rated residential mortgage
backed securities would have seen a tenfold increase in haircuts [from 2% to 20%],
meaning that its leverage must fall from 50 to just 5’ (emphasis added). Here is market
evidence of a systemic bank run of such dramatic proportions as to precipitate prompt
insolvency of the shadow banks16, absent official support.

The crisis account of the FCIC can, it seems, be encompassed by appropriate
reinterpretation of the basic framework, as in Fig. 4. When A, interpreted as a ‘fooling
equilibrium’ (where creditors are lulled into a false sense of security by inflated ratings)
comes to an end, with asset prices falling as indicated by B, then shadow bank portfolios
contract as their equity takes a hit, see point C. This generates visceral panic as creditors
wake up to the risks they are exposed to - by banks whose liability is strictly limited to

15
 where the ‘haircut’ is the difference between the current market price of the collateral and the
price at which it is sold to the creditor – for future repurchase.
16
 With leverage for the big five US investment banks of about 30 before the crisis, Miller et al.
(2018), Table 2, col. 5.
 9

the loss of their own ( now reduced) equity – leading to collapse of shadow banking (as
prices fall beyond what their equity can cover), as at D in the figure17.

To put it bluntly, the Congressional narrative involves what Akerlof and Shiller (2015) call
‘the economics of manipulation and deception’; with crisis as the denouement.

3.2 The Scholars’ Tale: of efficiency, productivity shocks and sunspots

 Before getting into detail, here is how shadow banking is viewed in the tale told by
these academics:

 In ”normal” times, the growth of the wholesale banking sector improves both
 efficiency and stability. Improved efficiency stems from the comparative
 advantage that wholesale banks having in managing certain types of loans.
 Improved stability arises because retail banks act as a buffer to absorb loans
 that wholesale banks sell off, in effect improving the liquidity of secondary loan
 markets.
 On the other hand, the growth of wholesale banking system makes the
 economy more vulnerable to a crisis. As occurred in practice, the high leverage
 of wholesale banks makes this sector susceptible to runs that can have highly
 disruptive effects on the economy. A contractionary disturbance that might
 otherwise lead to a moderate recession, can induce a run on the wholesale
 banking sector with devastating effects on the economy, as experienced during
 the Great Recession. Gertler et al. (2016)
The FCIC did not provide their own technical analysis; but the opposite is true in this
case. Three key papers relevant here run to many pages, and the number of equations
increases as the model is extended. (From 27 in Gertler and Kiyotaki (2015) – hereafter
GK(2015) - with a single bank sector, reaching 39 when wholesale and retail banks are
treated separately, Gertler et al.(2016), and more with the addition of a macro model in
Gertler et al.(2020)). Although the models with wholesale and retail banks treated
separately are more revealing, it is clearly impossible to do them justice in a short piece.
The key features of banking and its crises are common to all three papers, however, so
we focus on the earliest paper where, in any case, ‘the banking sector … corresponds
best to the shadow banking system which was at the epicentre of the financial instability
during the Great Recession’. GK(2015, p.2016).

Can the results the authors obtain be expressed in terms comparable with the two-
sector framework with which we started? In broad-brush terms they can: for here too
bank intermediation has to compete with ‘direct’ lending by households; and banks are,
likewise, so efficient that they could satisfy the needs of all borrowers were it not for the
limitation of an equity constraint. The workings of the model can, moreover, be
illustrated from the numerical results that the authors provide.

One key difference to note, however, is that here the equity constraint is not designed
to check portfolio risk-taking. The anticipated return on the liabilities of end-use

17
 Note that, with precipitous panic, equilibrium may shift promptly from B to D, bypassing C.
 10

borrowers is perceived as non–stochastic; so, except for unanticipated shocks, lending
involves no risk! The competitive edge of bank intermediation lies not in its capacity to
take on risk but in the assistance it can provide to enable borrowers achieve potential
gains in productivity. Banks can do this with unflagging efficiency, but for non-banks
(households) there are ‘management costs’ reflecting their lack of expertise in screening
and monitoring investment projects; and these costs increase more than in proportion
with the amount of assets being handled in the non-bank sector.

This is illustrated in Fig. 5 below, based on the reported calibrations, with expected
productivity growth of assets (‘capital’) of all borrowers measured in percent per annum
on the vertical axis. The productivity payoff for those holding bank-intermediated assets
is measured from the left hand axis (as in Figs. 2 and 4), and for household-managed
assets measured from the right. With bank intermediation, all borrowers could expect to
achieve the national rate of productivity increase – whose steady state value is
calibrated at 5% p.a. in the simulations. While a few households can match this on their
own, efficiency diminishes as non-bank handling of assets expands, as indicated by the
schedule HH, with calibrations suggesting the least efficient household will achieve less
than half potential productivity growth that banks can deliver.

In this model of riskless intertemporal optimisation - unlike the basic model of section 2
where perceived asset risk gives rise to liquidity preference - shadow banks pay
generous interest rates on their liabilities;18 and households discount asset yields by the
rate of time preference. So, as shown in Figure 5, the steady-state asset demands for
both sectors are shifted below expected payoffs by the rate of interest paid on ‘deposits’,
which matches the rate of time preference. As bank demand is also subject to a net
worth constraint19, the pre-crisis equilibrium involves direct financing, as shown at A in
the figure, where the asset price is unity and banks take about two thirds of the market.

18
 In the simulations, 4% p.a. is the steady state value of interest paid to bank creditors; in order
that money be held in a setting with no liquidity preference, it matches the household rate of
time preference .
19
 The equity constraint takes the form ≤ , where is the asset price, is holding by
banks, whose net worth is N .

 11

Asset price, Q
 & productivity
 N Net worth
 growth
 constraint
 Efficient Banks
 1.05
 H Interest
 Steady state
 paid by
 productivity
 H banks,
 increase,
 4 % p.a.
 5% p.a. A
 1.0
 B

 N
 0.95
 C
 Not
 strictly to
 scale

 0.85 D

 ℎ
 Stock of
 capital
 1 assets

Fig. 5. Key elements of Gertler and Kiyotaki (2015): a productivity shock followed by a
bank run

If there are no foreseeable investment risks to be considered, why the equity
constraint? The rationale provided is the danger of bankers misbehaving – not by getting
the assets they manage falsely rated, but by walking away with them! Thus, in this
narrative, the equity share of funding required by the market is not to prevent bankers
exploiting the convexity of payoffs with limited liability, but to check the simple venality
of those managing other people’s money, GK(2015, p.2019). As dishonest bankers will
soon get found out (and depositors ‘force the intermediary into bankruptcy at the
beginning of the next period’), the banker will lose the equity (the franchise value) that
will fund his or her retirement. Willingness to play straight can credibly be signalled,
however, by providing an appropriate fraction of funding from own resources – i.e. by
the banker putting enough ‘skin in the game’. How much equity is required for this
purpose is revealed by the leverage ratios actually observed, judged to have done the
job.

 12

An unanticipated productivity shock

Evidently, the shock that leads to crisis cannot be a widening of the perceived risk of
returns on lending – ex ante, no risk is perceived. Here the shock is an unanticipated,
nationwide fall in the rate of productivity growth, , the payoff expected from bank
lending. The fall is calibrated as ‘a negative 5 percent shock to productivity ’ – i.e. an
aggregate ‘technology’ shock that reduces growth achievable at an annual rate by a
modest quarter of one percent (as 0.05x0.05 = 0.0025). It is assumed, moreover, that
this will be reversed over time as the technology variable follows a deterministic,
autoregressive recovery process after the shock.

In circumstances described, the effects that follow are the results of a small, zero-
probability technology shock. What are these effects? As calibrated, the productivity
shock - powerfully amplified due to bank leverage - has a substantial impact on the
aggregated shadow bank balance sheets; calibration shows the asset price falling by 5
percent, as indicated by the point C; and - with bank equity apparently halved - the
banks’ share of the market, , falling by about a quarter20. (Why the fall in the asset
price is much greater than indicated by the slope of steady-state household demand
schedule is presumably because of higher discount rates in the credit crunch.)

A ‘bank run’

Even though the productivity shock is expected die away over time (with no repeat
anticipated), this is not the end of the story. For, given the effects shown at point C,
including the large drop in bank equity, a bank run becomes feasible - and in the
calibration that is what occurs. So, as in Cole and Kehoe’s (2000) account of self-fulfilling
sovereign debt crises, with a sunspot there is a systemic bank run, where banks are
wiped out and their assets transferred to the other sector, shown as the move from C to
D in the figure.

The part played by the productivity shock is to fulfil the technical condition for a run,
namely that bank assets - evaluated after wholesale transfer to household management
– are worth less than outstanding credit. So it is not, as Bernanke put it, that ‘bad news
leads investors to believe that the “safe” short-term assets they have been holding may
not, in fact, be entirely safe’: for in this narrative all assets are viewed as riskless! It is,
rather, that the shock increases the potential consequences of creditor coordination
failure.

Summary

The narrative here is of financial intermediation leading to greater efficiency in
managing capital assets, but a dangerous lack of resilience. The only factor that prevents

20
 In the absence of figures for the impact effect of the productivity shock before assets change
hands, a point B has been inserted showing an initial price fall akin to the size of the shock.
 13

banks from taking all business from households, indeed, is the ‘friction’ of an equity
component in their finance needed to prevent corporate theft. Bankers do not
misbehave; but they are nevertheless exposed to the vagaries of technology elsewhere
in the economy.

In Gertler et al. (2016), where shadow and commercial banks are treated separately
(and referred to as wholesale and retail respectively), the authors conclude:

 Another important area for further investigation involves the modeling of the
 growth of wholesale banking. Our approach was to treat this growth as the
 product of innovation as captured by a reduction in the agency friction in inter-
 bank lending markets. Among the factors we had in mind that motivate this
 reduction is technological improvements that permit less costly monitoring, such
 as the development of asset-backed securities and repo lending.
Thus, in this narrative, the expansion of shadow banking before the crisis, far from being
a move to a ‘fooling’ equilibrium, is seen as the fruit of efficiency-enhancing
technological and product innovation in banking – with the benefits snatched away in a
crisis brought on by extraneous shocks.

4. The Covid connexion?

What light might Covid throw on the relative plausibility of these two narratives? One a
highly-technical professional assessment that shadow banks were delivering efficient
intermediation services, until borrowers are hit by a zero-probability productivity shock
and creditors proceed to panic. The other from a broad-based Congressional Committee
sceptical of self-regulation and open to the idea of market manipulation – particularly by
shadow banks and rating agencies which had recently moved to limit their liability (with
Goldman Sachs going public in 1999, for example, and Moody’s in 2000.)

Four aspects of the current pandemic seem relevant here: how it began from a small
biological shock; how it gets spread by contagion; the significance of externalities; and
how it may end with a vaccine.

Small, unanticipated shocks with large effects.

Some say that a civet cat in a wet market in Wuhan may have provided the Covid virus
with a link from bats to humans – leading to devastating effects on health and the
economy across the world. Who knows? But if such a small, random event could
conceivably have such large effects, does this not lend substantial support to a narrative
that attributes financial crisis to a small, unanticipated productivity shock?

A lot depends on what is covered by the term productivity shock. For shadow banks, it
actually refers to unexpectedly bad portfolio payoffs. But, well before the financial crisis,
Rajan (2005) had warned that financial developments might be making the world riskier

 14

- as investors took on ‘tail risk’21 in particular. If so, such portfolio shocks could well be
endogenous - the result, perhaps, of portfolio managers without special skills mimicking
the returns of ‘alpha’ traders and paying for this by taking on low probability risk of
catastrophic loss.22

On one interpretation, therefore, the productivity shock could refer to a nationwide
hiccup in technical progress; on another, it could match the Congressional narrative of
hidden investment risk, with banks effectively following a strategy to ‘get these
securities rated high and quit when you’re hit’23.

The process of contagion

If the convenient but heroic assumption of Common Knowledge is dropped (as in the
FCIC narrative), where do market participants get their information from? The Credit
Rating Agencies have been discussed as one - none too reliable - source. Could it be that
ideas are also disseminated in a process of ‘social contagion’ rather like a virus, as
suggested by Robert Shiller when writing about the subprime crisis in 2008. What
relevance might this have to the crisis in shadow banking?

According to Shiller (2008, p.29) “The housing bubble was a major cause, if not the
cause, of the subprime crisis… The perception that real estate prices could only go up,
year after year, established an atmosphere that invited lenders and financial institutions
to loosen their standards and risk default.”

The logic is that ‘social contagion’ helped fuel a bubble in house prices; and this
underpinned the boom in subprime lending, where borrowers were encouraged to
share in the expected capital gains and the CRAs helped to conceal the risk that was
involved. As Holmstrom (2009, p.267) pointed out, however, the ‘dynamic credit
enhancement’ that underpinned subprime lending could only work if house prices --
already buoyed by a bubble -- continued to rise incessantly: so the subprime mortgage
boom was ultimately not sustainable. When house prices stopped rising, this, indeed,
had a devastating impact on the value of securities backed by mortgages issued during
the bubble. Fig. 6 shows the prompt collapse in the price indices for AA and BBB
packages of such securities (by 80% and over 90% respectively, rh scale) after house
prices had peaked – and how this impacted a broad index of bank shares.

21
 Where the probability that an investment payoff will move more than three standard
deviations from the mean is greater than that of a normal distribution.
22
 In a setting of asymmetric information, such downside risk in investment portfolios can- by
using derivatives - be concealed from outside observers for considerable periods of time, Foster
and Young (2010), Sinn (2010).
23
 “The names of the derivatives being used included IBG (I’ll Be Gone if it doesn’t work), and in
Chicago, the O’Hare Option (buy a ticket departing from O’Hare International Airport: if the
strategy fails, use it; if the strategy succeeds, tear up the ticket and return to the office). That
such strategies were common enough in the industry as to have names suggests that not all
traders were oblivious of the risks they were taking.” Rajan (2010, p.139).
 15

Fig. 6. US House prices, ABX indices, and share prices of Global Banks. Source: A. Milne
(2009, p. 201)

Both in his address on “Narrative Economics” to the American Economic Association and
in his monograph with the same title, Shiller (2017 and 2019) refers to the SIR model of
Kermack and McKendrick (1927) as one that might be applied to social contagion. As it
happens, however, a variant of the SIR model has already been applied to the housing
market using the Case-Shiller indices of house prices, namely Burnside, Eichenbaum and
Rebelo (2016), hereafter BER24. For the convenience of the reader, a brief outline of the
SIR epidemic model and a summary of the BER approach – along with a ‘health warning’
- are provided in the Annex.

Thus, while not explicitly acknowledged by Shiller himself, an account resembling that
being used for the Covid crisis25 has been used to explain the ‘epidemic of social
contagion’ in the housing market. This social epidemic - together with the disingenuous
role of the CRAs – could have played a key role in explaining how the financial system
got to what has been called the ‘fooling equilibrium’ of the FCIC narrative.

Significant externalities in play

24
 For this key reference, not to be found in Shiller’s own literature surveys, I am indebted to
Marija Vukotic.
25
 For the SIR model has, of course, been widely applied in studying the course Covid-19 - with
Ferguson et al. (2020) and Eichenbaum et al.(2020) being prominent examples from the UK and
USA.

 16

As each infected person can pass the virus on to others – to three others in the early
days of the UK first wave, for example - Covid involves significant health externalities.
Even when private action to avoid infection is allowed for 26, as it fails fully to take
account of the health of others, there is need for public policy to limit the externality
(with interventions to mandate face-coverings, social distancing, lockdowns, etc.) until a
vaccine is found. That national lockdowns have been imposed – and re-imposed – in so
many countries is testimony of how far policy-makers are willing to go to check the
externality of contagious spread.

If it encourages risk-shifting by banks with limited liability, the drive to maximise the
ROE will also generate significant externalities - in the form of unanticipated losses to
creditors and disruption to borrowers27. Subsequent to the crisis of 2008-9, significant
fines have been imposed on banks (for misleading others as to the quality of the MBS
they sold) and on CRAs (for deliberate over-rating).28 Such fines undoubtedly hit the
ROE, but getting corporations to internalise social costs appears to require targeting key
decision-makers, rather than shareholders in general29.

As Paul Romer (2012) argued in the wake of the financial crisis, however:

 There are workable alternatives to the legalistic, process-oriented approach that
 characterizes the current financial regulatory system in the United States. These
 alternatives30 give individuals responsibility for making decisions and hold them
 accountable.
In this spirit, Goodhart and Lastra (2020) propose that key decision-makers in banks and
significant financial institutions be made personally liable for downside losses ‘so as to
shift the costs of failure back to those who have responsibility for taking corporate
decisions’. Zingales (2020) argues likewise that a fiduciary duty towards society be
imposed (on all large corporations, not just big banks) with the board personally
responsible for damaging externalities.

Such proposals appear more consistent with the sceptical assessment of the conduct of
limited liability corporations made by the FCIC, than with the endorsement of their
financial innovations offered by G&K, as cited above.

The role of vaccine: is there an analogue available for ‘social contagion’?

26
 As in Eichenbaum et al. (2020), for example, where individuals work and shop less.
27
 Aikman et al.(2015) describe three ‘ games bankers play’ that can generate these externalities.
28
 Details in Miller et al.(2018, pp.103,104).
29
 Why the legal settlements have, nevertheless, taken the form of ‘deferred prosecution
agreements’ with the companies involved, rather than the criminal prosecution of high-level
individuals, is discussed in Rakoff (2014) - with the contrast between the US and Iceland in this
respect high-lighted in Miller(2019).

30
 The approaches to safety at the Federal Aviation Authority, to macroeconomic stabilization at
the Fed and to race relations in the army, are three examples cited.
 17

Vaccines - now being approved for general release against Covid-19 - promise an
accelerated transition to ‘herd immunity’, as shown graphically for the SIR model in the
Annex. This prompts the question: is there an analogue for epidemics of social
contagion?

For BER, where the bubbles arise from uncertainty (as to whether renting is preferable
to owning a house), the answer lies in the resolution of this uncertainty. How this might
be achieved is not discussed, however. In a recent Bank of England paper, Haldane et
al.(2020), epidemic models of social dynamics like BER are seen as cases where ill-
informed ‘noise’ prevails; and the alternative proposed is one of ‘narrative
entrepreneurship’ – of helping people make better sense of the economy and to form
more reasonable expectations. Would it be stretching their argument too far to
conclude that the Central Bank could supply the vaccine - reducing the ‘noise’ by
appropriate sectoral intervention on credit under macro-prudential regulation, for
example?

5. Conclusion

In remarks on the Global Financial Crisis, Jon Danielsson (2019, p.258) makes a useful
distinction between exogenous and endogenous risk. The former, as its name suggests,
comes from outside, affecting agents within the system without being influenced by
them. ‘Endogenous risk, on the other hand, is created by the interaction of economic
agents, all with their own agendas, abilities, resources and biases.’

In the narrative told by the scholars studied here, attention focuses on the vulnerability
of the seemingly-efficient ‘supply chain’ of finance in the face of exogenous
disturbances, be they unanticipated technology shocks (that damage bank portfolios) or
‘sunspots’ (that cause creditors to run). Endogenous risk seems effectively ruled out
however, with assets held by banks perceived as riskless and self-regulation keeping
managers honest.

At first blush, the Covid epidemic seems to support the idea of extraneous shocks being
the driver of crisis. Yet, in the FCIC telling, the banking crisis is like what happens in
Shakespeare’s Tempest, where it emerges that the fierce storm that opens the play was
called up by Prospero, one of the players! For, in the Congressional narrative, it is the
behaviour of the players themselves that generates much of the action. There are CRAs
ready to mis-rate risky assets for a fee; shadow bankers seeing this as a way of shifting
risk onto creditors despite the VaR rules in place; and creditors lending freely - until they
find out what’s happening and respond with swingeing haircuts to bank collateral!
Minsky famously argued that capitalist economies are unstable because of their financial
markets and institutions. This narrative supports no such universal proposition: but it

 18

does sound a clear note of warning - that capitalist economies can be destabilised when
financial markets and institutions put profits before purpose. 31

Here, as in Rubinstein’s fable, the choice of narrative is left to the reader. But we can
conclude on a positive note. For it appears that regulatory reaction to the near-collapse
of shadow banking - by making the system more resilient to risk whatever its source –
has put banks in better shape to handle the subsequent Covid pandemic!

References

Adrian, T. and H.S. Shin (2014) “Pro-cyclical Leverage and Value-at-Risk”. Review of Financial
Studies, 27 (2) 373-403.

Aikman, D., A. G, Haldane and B. D. Nelson (2015) “Curbing the Credit Cycle”. Economic Journal,
Vol 125(585) pp. 1072-1109

Akerlof, G. A. and R. J. Shiller (2015) Phishing for Phools: The Economics of Manipulation and
Deception. Princeton, NJ: Princeton University Press.

Bernanke, B. S. (2018) “Financial panic and credit disruptions in the 2007-09 crisis”. Brookings
Institution https://www.brookings.edu/blog/ben-bernanke/2018/09/13/financial-panic-and-
credit-disruptions-in-the-2007-09-crisis/

Bernanke, B. S, T. F. Geithner and H.M. Paulson (2019) Firefighting: the financial crisis and its
lessons. NY, NY: Penguin Random House.

Bolton, P. and M. Dewatripont (2005), Contract Theory. Cambridge, MA: MIT Press.

Burnside, C., M.S. Eichenbaum and S.Rebelo (2016). “Understanding Booms and Busts in Housing
Markets”. Journal of Political Economy, 124(4) pp. 1088-1147.

Cole, H. L., and T. J. Kehoe. 2000. “Self-Fulfilling Debt Crises.” Review of Economic Studies, 67 (1):
91–116.

Danielsson, J. (2019) “Financial Policy After the Crisis”. In: R. Z. Aliber and G. Zoega (eds.) The
2008 Global Financial Crisis in Retrospect. London: Palgrave Macmillan pp. 257-279 (Chapter 14)

Easley, D. & J. Kleinberg (2010) Networks, Crowds, and Markets: Reasoning about a Highly
Connected World. Cambridge: Cambridge University Press.

Eichenbaum, M. S., S. Rebelo and M. Trabant (2020). “The Macroeconomics of Epidemics”. NBER
Working Paper 26882

Financial Crisis Inquiry Commission (FCIC) (2011) Final Report. NY, NY: Public Affairs

Foster, D. P. and Young, P. (2010), “Gaming Performance Fees by Portfolio Managers”. The
Quarterly Journal of Economics, 125(4), 1435-1458.

Ferguson, N.M. et al.(2020) “Impact of Non-pharmaceutical Interventions (NPIs) to Reduce
COVID-19 Mortality and Healthcare Demand,” manuscript, Imperial College, March.
DOI: 10.25561/77482

Gertler, M. and N. Kiyotaki (2015) “Banking, Liquidity, and Bank Runs in an Infinite Horizon
Economy”. American Economic Review, 105(7): 2011–2043

31
 The Icelandic banking collapse of 2008 provides a fascinating supporting act – described in
damning detail by Gudrun Johnsen (2014), using the findings of Iceland’s Special Investigation
Commission to do so.
 19

Gertler, M., N. Kiyotaki and A. Prestipino (2020). "A Macroeconomic Model with Financial
Panics," Review of Economic Studies, 87(1): 240-288.

Gertler, M., N. Kiyotaki and A. Prestipino (2016). “Wholesale Banking and Bank Runs in
Macroeconomic Modeling”. In: Taylor, J.B., and H. Uhlig. (Eds.), Handbook of Macroeconomics.
Vol. 2B: Elsevier, Amsterdam, Netherlands. Pp. 1345-1425.

Goodhart, C. A. E. and R. M. Lastra (2020) “Equity Finance: Matching Liability to Power”. Journal
of Financial Regulation. 6(1) pp. 1-40. Also circulated as CEPR DP No. 13494

Haldane, A., A. Macaulay and M. McMahon. (2020) “The 3 E’s of central bank communication
with the public”. Staff Working Paper No. 847, Bank of England.

Holmstrom, B. (2009) Commentary: “The Panic of 2007”. Maintaining Stability in a Changing
Financial System. Kansas City, MO: Federal Reserve Bank of Kansas City.

Johnsen, G. (2014) Bringing down the banking system: lessons from Iceland. Basingstoke:
Palgrave Macmillan

Kermack, W. O. and A. G. McKendrick. 1927. “A Contribution to the Mathematical Theory of
Epidemics.” Proceedings of the Royal Society. 115 (772):701-21

Miller, M. and L. Zhang (2019) “Externalities and financial crisis – enough to cause collapse?”
CEPR DP 13834.

Miller, M. (2019) “The Financial Alchemy that Failed”. In: R. Z. Aliber and G. Zoega (eds.) The 2008
Global Financial Crisis in Retrospect. London: Palgrave Macmillan pp. 19-41 (Chapter 2)

Miller, M., S. Rastapana and L. Zhang (2018) “The blind monks and the elephant: contrasting
narratives of financial crisis.” The Manchester School Vol 86 No. S1 83–109 September

Milne, A. (2009). The Fall of the House of Credit: What Went Wrong in Banking and What Can Be
Done to Repair the Damage? Cambridge: Cambridge University Press.

Moll, B. (2020) “Lockdowns in SIR models.” Technical report, LSE (May)
https://benjaminmoll.com/wp-content/uploads/2020/05/SIR_notes.pdf

Rajan, R. G. (2010) Fault Lines. NJ: Princeton University Press

________ (2005) “Has Financial Development Made the World Riskier?” The Greenspan Era:

Lessons for the Future. Kansas City, MO: Federal Reserve Bank of Kansas City.

Rakoff, J. S. (2014). ‘The Financial Crisis: Why Have No High-Level Executives Been Prosecuted?’,
The New York Review of Books (January 9).

Romer, P. (2012), ‘Process, Responsibility and Myron’s Law’. In: Blanchard et al. (eds.) In the
Wake of the Crisis: Leading Economists Reassess Policy. Cambridge, MA: MIT Press. pp. 111-121
(Chapter 12)

Rubinstein, A. (2012) Economic Fables. Cambridge, UK: Open Book Publishers

Shiller, R.J. (2019) Narrative Economics. Princeton, NJ: Princeton University Press.

________ (2017) “Narrative Economics.” American Economic Review, 107(4), pp.967-1004.

________ (2008) The Subprime Solution. Princeton, NJ: Princeton University Press.

Shin, H. S. (2010) Risk and Liquidity. Oxford: Oxford University Press

Sinn, H. W. (2010) Casino Capitalism: How the Financial Crisis Came About and What Needs to be
Done Now. Oxford: Oxford University Press.

Smith, D. and L. Moore (2001) “The SIR Model for Spread of Disease”. Journal of Online
Mathematics and its Applications. https://www.maa.org/press/periodicals/loci/joma/the-sir-
model-for-spread-of-disease.
 20

Zingales, L. (2020) “Which Capitalism?” Presentation at Markus’ Academy Webinar, Princeton
Bendheim Center (Oct 8)

Annex: Medical and Social epidemics

The SIR Model for Spread of Disease (Smith and Moore,2001)

There are three groups in the population, Susceptible, S, Infectious, I, and Recovered, R,
which evolve as follows:

 ̇ = −β (1)
 ̇ = β − γ (2)

 ̇ = γ (3)
where β is the number of close contacts per period between susceptibles and those
already infected; γ measures the rate at which the infected population ceases to be
infectious; so β/γ is the ‘reproduction ratio’ - the initial rate of spread; and the
constant population is normalised at 1.

The dynamics of S and I are shown in the phase diagram below, along with the peak
level of Infected population (P),the Herd Immunity Threshold (HIT) and the specific level
of Herd Immunity ( ∞ ) where infections cease in the long run with no intervention. The
parameters used are based on the Hong Kong flu epidemic in New York City in the late
1960's.

 21

S, Susceptible

 Not to scale

 UNSTABLE

 1

 S= γ/β = 2/3 ̇ = 0 Herd Immunity Threshold (HIT)
 (with β =1/2, γ =1/3)

 0.4 ∞ Herd
 STABLE
 Immunity

 P peak of I
 0.07 1 I, Infectious

Fig. A1. Epidemic unchecked - except by vaccine

No indications are given here for non-pharmaceutical interventions such as social
distancing, lockdowns, and test/trace /isolate measures32; but the dashed line illustrates
the impact of a vaccine for about half the population that takes the susceptible
population promptly below the Herd Immunity Threshold, without affecting the level,
 ∞ , where infections finally cease. (More or less vaccination would affect this level.)

The Social Dynamics of BER, i.e. Burnside, Eichenbaum and Rebelo (2016)

BER use a version of SIR, but it is considerably modified as follows:

All agents have identical tastes, but hold different beliefs about a possible change in the
utility of home owning versus renting. The three groups are Vulnerable, Skeptical and
Optimist, where the latter believe with confidence that the utility is going to increase by
a factor x; while the other two believe there will be no change, of this Skeptics being
more sure than the Vulnerable. When members of any group meet those of another,
beliefs can be affected: in particular, those with greater confidence can persuade others
to adopt their beliefs. Unlike the SIR model, where the transition from S is to I and then

32
 For which Moll (2020) provides an excellent overview, with parameters more appropriate to
the current pandemic.
 22

to R , in BER there is greater mingling; thus one can transfer from Vulnerable to either of
the other two groups; and there is also switching between the groups.

Depending on parametric values of these switching probabilities, starting with very few
of either Optimist or Skeptical, one can expect a temporary boom in house prices
followed by a reversal as Skeptical beliefs prevail; or a boom that is not reversed where
Optimistic beliefs prevail. This is when the uncertainty is unresolved. If it is resolved in
favour of the increase x, house prices will go up to boom levels; if not then prices will
revert to original values.

Easley and Kleinberg (2010, chapter 21) have issued a health warning about using
biological models to describe social contagion, on the grounds that the latter involve
conscious decision-making while the former are based on random transmission. Is the
notion of ‘strength of belief’ used by BER to govern the transmission of ideas an
adequate response?

 23